Live event transcription for 3 hours: real-time vs batch cost
Three hours of real-time transcription is $1.17 on ElevenLabs Scribe Realtime and $0.60 on xAI streaming; Sume's batch STT is $1.80 after the event.
Transcribing a 3-hour event live costs $1.17 with ElevenLabs Scribe v2 Realtime at $0.39 per hour and $0.60 with xAI streaming at $0.20 per hour. Transcribing the recording afterward costs less on the batch tiers: $0.66 on ElevenLabs Scribe v2 ($0.22 per hour) and $0.30 on xAI REST ($0.10 per hour), against $1.80 on Sume at $0.01 per audio minute.
Prices are from the ElevenLabs and xAI pages (read 2026-10-04) and Sume's API pricing. Sume's STT is a batch job, so it does not give you live captions.
| Option | Rate | Calculation | Total |
|---|---|---|---|
| ElevenLabs Scribe v2 Realtime | $0.39/h | 3 x $0.39 | $1.17 |
| xAI streaming | $0.20/h | 3 x $0.20 | $0.60 |
| ElevenLabs Scribe v2 (batch) | $0.22/h | 3 x $0.22 | $0.66 |
| xAI REST (batch) | $0.10/h | 3 x $0.10 | $0.30 |
| Sume STT (batch) | $0.01/min | 180 x $0.01 | $1.80 |
Live or after the fact
Real-time pricing buys captions while people are speaking. Realtime costs 1.77x ElevenLabs batch ($0.39 / $0.22) and 2.0x xAI batch ($0.20 / $0.10). If you only need a transcript for the minutes or the clips, the batch rows are cheaper for the same audio.
Sume's speech-to-text is a batch job with a 10-minute cap per audio file, so a 3-hour recording is 18 jobs of 10 minutes at $0.10 each, 18 x $0.10 = $1.80. A failed slice is refunded and only that slice is resubmitted.
- Live captions, subtitles for the room: a real-time tier.
- Minutes, summaries and clip hunting: batch on the recording.
- Recording plus clips: transcribe in Sume, then use the text to choose cut points.
Steps for the batch route
- Record the event audio to a file.
- Split it into 10-minute slices at pauses, keeping the offsets.
- Submit the 18 slices and merge the transcripts by offset.
- Use the merged text to pick segments for video trims, listed at $0.02 per job.
A year of events
Twelve three-hour events a year is 36 hours. Real-time costs 36 x $0.39 = $14.04 on ElevenLabs and 36 x $0.20 = $7.20 on xAI. Batch on recordings costs 36 x $0.22 = $7.92 on ElevenLabs, 36 x $0.10 = $3.60 on xAI and 2,160 x $0.01 = $21.60 on Sume. Even the highest figure is small next to the cost of a venue or a crew.
Caveats
Pricing aside, accuracy on a live room with crosstalk varies by model, and real-time services add latency and reconnect handling that batch does not. Rates read 2026-10-04.
Sources
Related posts
More in Comparisons
- How long can an AI song be? Suno, ElevenLabs and Lyria compared
ElevenLabs Music runs 3 s to 5 minutes; Suno v6 is reported at up to 8 minutes; Lyria 3.5 makes a couple of minutes. Join tracks with a $0.01 concat.
- LTX-2.5 license ($10M ARR line) vs hosted video ids
LTX-2.x Community license is free under $10M ARR and paid above it. Here is what changes if you call a hosted Sume video id instead of running weights.
- LTX-2.5 Fast 20 s vs Pro 10 s: long takes on Sume
LTX-2.5 Fast runs up to 20 s at 720p and 1080p, Pro 6 to 10 s. Sume lists no LTX id, but wan-3.0 and seedance-2.5 take single clips up to 30 s.
- Luma API callbacks and credit balance vs Sume webhooks and /v1/balance
Luma's docs list callbacks and a credits balance. Sume has webhook mode, GET /v1/balance, and a generation_limits snapshot. How to use each before a batch.
Written by Sume